Autism is characterized by deficits in communication, social interactions, and unusual interests and stereotyped behaviors. Individuals with autism often perform better on tasks with reduced levels of perceptual complexity. This proposal will investigate how children with autism interact with computer-generated avatars with multiple levels of physical detail and movement characteristics to determine the effects of these modifications on social interactions. The main hypothesis of the proposed studies is that the social deficits observed in children with autism will be ameliorated during interactions with characters with reduced levels of detail. Specifically, we will build upon the existing work to use avatars to test the effects of modifications in appearance and motion complexity on conversation behaviors. Laboratory studies will include verbal children with autism between the ages of six and ten years. A new, state-of-the-art technology will be used that allows for live interactions and spontaneous conversations between the participant and an actor with various appearances. This research will be the first to compare the behaviors of children with autism while they interact with multiple types of avatars and among the first to use live, unstructured interactions with unconstrained scripts between avatars and children with autism. The avatars will vary in appearance complexity as well as motion complexity and the amount of visual and auditory complexity presented over time. This work will increase knowledge about basic perceptual processes in autism and impact future interventions with this population. The proposed research will build upon recent progress in animation methods, cognitive science, developmental science, and human-computer interaction research. Moreover, it will have implications for treatment practices with computer-based interventions for children with autism by informing investigators about the appropriate selection and use of various avatars.